Development of a smart tomato sorting system using naïve bayes algorithm based on color, size, and weight

Post-harvest sorting plays a crucial role in ensuring product quality and market consistency, yet manual tomato grading often suffers from low accuracy and inconsistent results due to human subjectivity. To address this issue, this study proposes an automated tomato sorting system that integrates multi-sensor data with the Naïve Bayes classification algorithm as an efficient alternative to manual sorting. The research contribution is the development and experimental validation of a low-cost, physics-based decision support system that fuses color, size, and weight information for tomato quality classification. The system employs a TCS230 color sensor, an ultrasonic sensor for size estimation, and a load-cell sensor for weight measurement, with all features processed using the Naïve Bayes algorithm to categorize tomatoes into Class 1 (premium), Class 2 (medium), and Class 3 (standard). Experimental results show that the system achieves an accuracy of approximately 92%, with strong performance for clearly distinguishable classes and reduced performance for tomatoes with irregular coloration or shape variations. These findings demonstrate the feasibility of probabilistic sensor fusion for improving post-harvest tomato sorting, and further improvements may include expanded datasets, enhanced calibration, and integration of computer-vision features for increased robustness.